Local Sensitive Frontier Analysis based facial expression recognition

Facial expression recognition plays an important role in interactive entertainment. In this paper, LSFA (Local Sensitive Frontier Analysis) a novel feature extraction method is introduced for facial expression recognition. LSFA is designed as manifold based feature extraction method to obtain useful features from the facial expression pictures, since the facial expression scatter in high dimensional space as a point will embed in low dimensional manifold. From comparing several feature extraction methods in the experiment, it can be found that this algorithm gets better expression recognition result.